Could CIDER provide an xref backend instead of hijacking M-. and M-,?

[stig]

Is your feature request related to a problem? Please describe.

I'm frustrated with the way CIDER hijacks M-. in all clojure-mode buffers, even for projects where you don't connect to CIDER. I don't always want to connect to CIDER to navigate projects efficiently.

Describe the solution you'd like

I would love for CIDER to provide an xref backend instead of hijacking M-., similar to how clojure-lsp does.

My hope is that this would mean xref would present results both from CIDER and clojure-lsp, and other backends.

Describe alternatives you've considered

A simpler (though less ideal) fix would perhaps be to fall back to xref-find-definitions if there's no CIDER connection for the project.

Additional context

Add any other context or screenshots about the feature request here.

[bbatsov]

Yeah, we can certainly do this. CIDER predates the existence of xref in Emacs, which is the reason for the current state of affairs. I meant to play with xref a few times, but never got to doing so. It's mostly a matter of finding time.

[bbatsov]

@stig Btw, are there some packages implementing this outside of Emacs that I can take a look at? Or any documentation on the subject?

[stig]

I didn't find any documentation, but I appear to have three packages that implements it, according to documentation for xref-backend-definitions:

[...]
Implementations:

((backend (eql xref-lsp)) identifier) in ‘lsp-mode.el’.

Undocumented

((backend (eql etags)) symbol) in ‘etags.el’.

Undocumented

((backend (eql elisp)) identifier) in ‘elisp-mode.el’.

Undocumented

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contribution and understanding!

[bbatsov]

I've started work on this a while ago, but it turned out to be more involved than I anticipated and I've paused it for now. I plan to revisit this down the road, but anyone's welcome to beat me to it. :-)

[bbatsov]

Here's some useful reference pythonic-emacs/anaconda-mode@39b1cf8

[jiacai2050]

@bbatsov Any news for this?

[bbatsov]

No progress. I've been pretty busy lately.

[jiacai2050]

@bbatsov I would like to try submit a PR to fix this, maybe in two/three weeks.

[juszczakn]

This isn't working for me, for some reason. I keep getting the prompt Visit tags table (default TAGS): /some/path/...

I'm going to try debugging it, but I'm not really much of an elisp guru /shrug

Edit:
Ok, it seems like my xref-backend-functions is evaling to a list with t as the first value. Removing t seems to resolve the issue for me.

> xref-backend-functions
(t cider--xref-backend)

I'm temporarily fixing this in my init.el by just getting rid of the t with:

(add-hook 'cider-mode
          (lambda ()
            (setq xref-backend-functions '(cider--xref-backend))))

which is definitely not a good thing, I'm sure, but it's working for me, for now /shrug.

My guess is that Emacs is working through the list, finds t, and just assumes it should use the default, which is etags?
This is Emacs 26.3 on Ubuntu 20.04.

---

Edit2:

Hook ended up not really working very well, seems like upgrading to emacs 27.2 (using the ppa) has fixed it, though.